Introduction:

Amyloid-β (Aβ) positron emission tomography (PET) often fails to predict near-term cognitive decline, limiting its prognostic value in Alzheimer’s disease (AD).

In line with AD’s proposed pathophysiological cascade of an Aβ-triggered tau phosphorylation response, plasma p-tau217 is more closely linked to subsequent cognitive decline than Aβ burden.

Incorporating this cascade into disease progression models, rather than treating p-tau217 as an interchangeable proxy for Aβ, may therefore improve AD risk stratification.

 

Aim: to investigate the combined prognostic value of plasma p-tau217 and Aβ PET in predicting cognitive decline.   

Methods:

Multicohort longitudinal study: n = 2,902; 5 cohorts (ADNI, A4, HABS-HD, BICWALZS, and WRAP).

822 cognitively impaired and 2,080 cognitively unimpaired at baseline.

Baseline: Aβ PET Centiloid scale, plasma p-tau217 and clinical dementia rating scale (CDR).

Impairment defined as CDR sum of boxes (CDR-SB) > 0 and no impairment as CDR-SB = 0.

Follow-up (up to 6 years): progression defined as any increase in CDR-SB or worsening diagnosis.

Sensitivity analysis for Mild Cognitive Impairment (MCI) conversion to dementia, using Global CDR of 0.5 converting to 1.0.

Cox proportional hazards models and Kaplan-Meier analyses of progression-free survival.

Biomarker positivity thresholds: Aβ PET Centiloid ≥ 20; p-tau217 z-score ≥ 1.

Results:

Table 1. Hazard ratios and progression risk across all four biomarker profiles. Hazard ratios (HRs) with 95% confidence intervals (CIs) are shown across biomarker profiles, with arrows indicating values beyond the axis limits. P values are derived from Wald tests of model coefficients and Bonferroni-corrected within each group-specific Cox model across all reported coefficients. Models were adjusted for age, sex, and education, and stratified by study cohort. Absolute risk (AR) is also reported with 95% CIs. The double-positive profile was associated with a significantly higher risk of progression, whereas the profile positive for Aβ PET and negative for p-tau217 was not associated with increased risk and comprised a substantial proportion of Aβ PET-positive individuals (27.8% of impaired and 36.5% of unimpaired participants).

ΔC-index

We compared two nested Cox proportional hazards models (adjusted for age, sex, and education, and stratified by study cohort): a base model using continuous Centiloid values and an augmented model that added plasma p-tau217 values.

p-tau217 significantly improved discrimination: ΔC-index was 0.0206 (95% CI, 0.0078–0.0333; P = 0.0015) in the cognitively impaired group, 0.0480 (95% CI, 0.0253–0.0707; P < 0.0001) in the MCI-to-dementia sub-analysis, and 0.0259 (95% CI, 0.0147–0.0370; P < 0.0001) in the cognitively unimpaired group.


Figure 1. Progression-free survival curves across all four biomarker profiles. Kaplan–Meier estimates of progression-free survival across biomarker profiles defined by Aβ PET and plasma p-tau217 status in participants classified at baseline as (a) cognitively impaired and (b) cognitively unimpaired. Double-positive individuals (Aβ PET+/p-tau217+) show the shortest progression-free survival, whereas Aβ PET+/p-tau217− individuals show survival comparable to that ofthe double-negative group

Conclusion:

Plasma p-tau217 refines prognostic stratification beyond establishing Aβ positivity, even when amyloid burden is modeled continuously. Our results support a complementary, dual-biomarker framework in which Aβ PET confirms amyloid status and plasma p-tau217 adds clinically relevant prognostic information to support risk stratification and participant selection in clinical trials.